B2G/Meeting/2012-12-04

From MozillaWiki
< B2G‎ | Meeting
Jump to: navigation, search

News

Administrative

Ecosystem

  • <ateam> (mdas, jgriffin, ahal, wlach) - updated 12/04
    • This week:
    • Emulator unit tests:
      • xpcshell tests running on TBPL on cedar with a few problems; should make the Dec 10 deadline
      • work on adding more mochitests, reftests, and xpcshell tests will continue past Dec 10; this is conditional on addressing the linux32 slave issues in buildbot
      • There is an occasional b2g crash (https://bugzilla.mozilla.org/show_bug.cgi?id=814551), and an occasional emulator crash (https://bugzilla.mozilla.org/show_bug.cgi?id=802877), that are making sheriffing difficult for these tests; what can we do about these?
    • Gaia smoke tests on pandas
      • No major blocking Marionette issues, although there are several issues that affect individual tests or small groups of tests
      • Big issue is that we still can't connect to Marionette remotely (https://bugzilla.mozilla.org/show_bug.cgi?id=800138). We used to be able to get around this by loading Marionette later in the startup cycle, but this no longer works. Need help debugging!
    • Performance tests
      • Ted is working on a patch to allow us to get raw app load times on a panda board
      • Will is working on getting fps measurements with Eideticker, but is having a lot of problems with B2G stability on pandas. See https://bugzilla.mozilla.org/show_bug.cgi?id=817730.
    • All panda-related work is at risk for Dec 10.
    • We're also being hit with partner requests (e.g., https://bugzilla.mozilla.org/show_bug.cgi?id=816099) which is impacting the amount of time we have to spend on other goals
    • Last week:
    • rolled out a new stable update today
    • ahal is working on expanding set of reftests being run on TBPL
    • ahal is working on adding support for running xpcshell tests in TBPL
    • malini, myself and others continue to work on supporting WebQA's automation of gaia smoke tests by fixing Marionette bugs and implementing enhancements
    • several people are working with rel-eng to try and get gaia smoke tests in running on pandas by Dec 10
      • have pref based work around for bug 800138, still need to follow up on what the right fix should be to start marionette after final-ui-startup but running with system context
      • Have fix for pandas do not boot with network enabled
      • Mozpool ready for roll out.
  • <QA team> (Geo, Jason, Tony)

BLOCKERS

    • 7 B2G and Gaia blockers: http://bit.ly/TAV5A4
      • Packaged apps still broken out of process
    • awaiting mozilla packaged - Bug 815781 - Upload .zip for unagi/otoro builds containing all files required for flashing
      • Preloaded apps story is still ambigious

STATUS

    • verified Releng infrastructure for unagi/otoro single and multilocale builds
    • verified dogfood builds moving to beta channel
    • Script to help with unattended triage of mochitests mostly finished: http://people.mozilla.com/~gmealer/triage-mochi.bash
      • Walks mochitest tree
      • Runs each test singly, saving the result to a tree of logs mirroring the test tree
      • Kills the framework, cleans it up, moves to next test
      • Framework kill necessary because single test runs never exit in B2G mochitest: https://bugzilla.mozilla.org/show_bug.cgi?id=780034
      • Right now, requires manual review of the log files, but very quick to flip through them.
      • TODO: scan test results to auto-build candidate list. Still collecting failure patterns
    • Have already triaged /caps, /docshell, /content, found many candidates. Will be adding passing results to TBPL run; would like to run in parallel first if possible to make sure no oranges.
    • Processing /dom with this script this week. Takes ~3 days straight to run for that many testfiles, as test files take 1.5-2min each to run due to framework setup cost and there are a few thousand files.
    • Gave a first shot at modifying mochitest framework to centrally set test permissions. Loading method did not work; taking second shot this week. Would appreciate active dev support if contacted.
      • Concern - Weird payments feature with a certified API (WebSMS) for web payment providers through mozPay - should we even do this?
    • Last week:

BLOCKERS

    • Regressions, regressions, regressions. Smoketest bugs are still hovering around ~16 a day, with some still not being commented or resolved within 24 hours. What can we do to continue driving priority on fixing these?

STATUS

    • QA tested a tinderbox build for unagi
    • platform bugs are being bb-'d if its unagi only? (photo preview issue, busted geolocation-bug 815217)
    • Status on bug: https://bugzilla.mozilla.org/show_bug.cgi?id=812396
      • Can we get this in as part of the build process ?
    • Logging discussion from earlier this week: default to no logs affects only Gaia logs? What about dogfooders?
    • Continuing to manually triage existing mochitests for addition into the manifest running on TBPL. Have found a significant number in /layout, a few elsewhere. Some fruit, slow going.
    • Working on automated method to walk the mochitests, execute file-by-file, and find passing candidates for inclusion. Spidering tests is easy, but framework hangs at the end of single test runs so execution is a little more complicated due to that.
    • dclarke driving effort to get permissions handling for functional API testing centralized into the mochitest harness. A-Team is providing harness knowledge, need more help from dev for B2G knowledge (fabrice?): https://bugzilla.mozilla.org/show_bug.cgi?id=814140
    • Progress continues on permissions testing w/ Security Assurance. Patches starting to show up on Bugzilla. Suite 1 & 3: https://bugzilla.mozilla.org/show_bug.cgi?id=811141 and Suite 2: https://bugzilla.mozilla.org/show_bug.cgi?id=815105
    • Apps - supported defining preloaded apps scope, including the testing piece, burning down exploratory testing. Getting ramped up on privileged app testing.
  • <Release Engineering> (aki, joduinn, hwine) - Updated on Dec04
  • This week:
    • Builds+Updates
      • [pending response] full emulator builds in releng's hands by dec10? akeybl will respond, not needed by dec10
      • [done?] manifest publishing -- response to come in Releases Meeting email thread? dietrich hasn't heard anything
    • Multilocale
      • [Unassigned blocker] gaia profile black in multilocale builds https://bugzil.la/817170
        • Does Fabrice have enough information to debug this?
      • otoro, unagi, panda gaia multilocale builds blocked on https://bugzil.la/818032
        • Stas has a loaner buildslave to try to debug this
      • Axel landed gecko-multilocale build Makefile changes; pending review for packaging steps in https://bugzil.la/817197
        • Aki to verify this works, still an unknown
    • Pandas
      • [in progress] Need to update panda snapshot to fix image https://bugzil.la/818264
      • [landed, needs reconfig] panda builds against gaia-central https://bugzil.la/817811
      • Marionette workaround landed, didn't work; blocking remote testing https://bugzil.la/800138
        • "So currently, the Marionette socket never stays open if force-local is false, no matter how late we start it, even if we wait 5 minutes after boot. This seems to indicate that something has changed in the last few weeks, since this didn't used to be the case."
      • Need gaia smoketests zipped up as part of build https://bugzil.la/817040
      • Can't currently install builds from pvtbuilds onto Pandas: https://bugzilla.mozilla.org/show_bug.cgi?id=818329
      • [ready to land] sendchange from b2g_build.py
    • continue working down this https://etherpad.mozilla.org/b2g-builds
    • repos now mirrored onto git.m.o - consistent URL needed for handover to partners
      • being evaluated by partners
      • change to manifest to include ".git" as part of the location
  • Last week:
    • Builds+Updates
    • Multilocale
      • [tested] QA verified that Otoro multilocale doesn't [always] brick phones https://bugzil.la/817108
      • [landed] - Desktop multilocale builds
    • Pandas
      • [landed] mozpool mozharness integration https://bugzil.la/810439
      • [landed] IT fixed panda flashing with network https://bugzil.la/810045
      • [tested] Clint verified we could run Gaia smoketests on a Panda at all, though ADB only
      • [landed] dummy test builders for Pandas on Cedar https://bugzil.la/816237
      • [done] IT set up a staging Panda mozpool server for RelEng to test against
      • [done] gaia-nightly and gaia-central repos replicating every 10min
    • Emulator tests
      • XPCShell now on Cedar, can roll out to m-c etc. when A-Team is ready
      • Now highlighting b2g emulator crashes in logs
  • <product team> (Chris Lee)
    • This week:
      • v1.1 release is in the works with a short list of requirements
    • Last week:
      • User story audits happening this week.
      • We are meeting with TEF next week to review the overall program (Product, Engineering, UX, Marketing, Apps, BD, etc.)
      • Working with release management to prioritize all dogfood blockers to kick off the program asap
      • Coordinating with OEM partners on software updates
      • Working with the team on prioritizing remaining feature work
  • <PMO team> (Kevin Hu, Dietrich, Michael)
    • This week:
      • Lots of partner negotiations
      • Performance, visual design bug criteria
      • Multi-market deployment plans
  • <Release Management> (Alex Keybl, Lukas Blakk)
    • This week
      • continuing convergence milestone work, readying for C3
      • mozilla-b2g18 is coming, along with associated tracking/status/approval flags
      • blocking-b2g for partner work is coming (blocking-b2g:TEF+, for instance)
      • dogfooders are being moved to the 'beta' channel, expansion to remoties is incoming
    • Last week:
    • Dogfooding:
      • Pulled today's update since it was bricking phones http://bugzil.la/815895 -- will be gathering logs today/tomorrow to try and help with investigation
      • Preparing to expand 140 additional phones to remote employees (signup announce going out today)
        • Blockers to expansion: updates moving to 'beta' channel, LDAP-auth dir for promoted builds in case people need to access for flashing (ETA end of week)
      • This week and next data-driven re-assess of tester participation and update pings to keep phones in most active users' hands
      • Will be doing a push on flashing phones to post-FOTA update and getting people's phone numbers in Phonebook for SMS and dialer testing 'games' in coming weeks
    • Convergence:
      • C1 bugs have either been blocking-basecamp:-'d or moved into C2 at this point.
      • C1 landing process extended for another week. Gecko developers - you MUST land on m-c, m-a, and m-b.
      • C2 bugs take priority over everything but P1/Critical (which should already be C2). See http://bit.ly/QKdwGX for the full query.
      • Only a couple of weeks from branching Gecko and starting the mozilla-b2g18 repo
      • Continued discussions around B2G updates
  • Developer Documentation (sheppy)
    • This week:
      • not in attendence

User Documentation & Support

(Michelle has a conflict today, just read the wiki)

  • Dogfooding and sending feedback summaries
  • Mobile redeisgn is landing and looking good
  • Action on the in-product links to support from Settings, thanks for the help there
  • Eveything.me will contribute help articles and forum support to SUMO per verbal agreement last week
  • TEF cancelled our meeting this week, so no new SLA progress
  • Starting training documentation outlines and scheduling
  • Based on 'second freeze' features coming in, I'd like to plan on conducting doc reviews during the last weeks of November, so we can incorporate all changes in early Dec.

Status updates

Gaia

APIs/Platform Work

  • Payments+Identity (Ferjm, Badida, Jedp, Caitlyn)
    • 5 more on-device bugs to fix before ready for e2e testing
    • waiting on eta from caitlyn
  • Camera (Mike Habicher) - Updated on 12/04
    • This Week:
      • bug 817141 - bb+ - r+: camera preview doesn't work from lock-screen
      • bug 817496 - bb+ - r+: no DeviceStroage onChange event when video recorder finishes
    • Last Week:
      • bug 809259 - bb+ - LANDED: crash when rapidly cycling between video/picture modes
      • bug 816817 - bb+ - LANDED: improper clean-up of MediaStreams
    • In The Queue:
      • bug 795332: "record-size" option may be deprecated in favour of "video-size"
        • not a big issue for Otoro, since the camera hw only supports one video stream, and its properties come from "preview-size"
      • bug 789067 - PATCH READY: clean up extraneous abstract ICameraControl class
        • patch posted to bug, still waiting on (increasingly obselete :) review
      • bug 786756 - PATCH READY: exposing more camera frobs
      • bug 813685: lag between camera and viewfinder
        • about 1/3rd of a second lag between moving the phone and updating the viewfinder
  • RIL: Cell broadcast (Vicamo) - Updated on 12/4
  • Bug 778093: B2G RIL: support Cell Broadcast
    • All review+. Vicamo already checked the codes into inbound.

Roundtable

Action Items

   joduinn/jgriffin: "addressing the linux32 test slave load issues in buildbot"?
   dietrich: contact t-zimmerman about panda resources and a-team blockage